In the last year I have had the privilege of meeting lots of companies that take advantage of the public IaaS capabilities. While most of the companies are already using Amazon AWS, unfortunately, I see many IT leaders that are having bad experiences with Amazon AWS resources sprawling out across their organizations, resulting in increased uncertainty of the monthly cost. It's a fact that cost is becoming a sore point in the cloud adoption phase, so instead of opting for a smart and planned process, the bad experience leads the IT leader to return to the "good old" "old and good" environment and simply buy new servers.
It seems that the decision makers have accepted that cloud computing leads to lower IT cost. Although many experience savings while switching to the cloud, the total cost of ownership for most is still too high. So before you go and spend more money on additional monitoring and management systems, here are some tips to help you regain control and reduce costs immediately.
Delegate Responsibility - Assign a dedicated person to control the "faucet" In other words don’t think that on top of your regular daily workload, you will have the time to maintain and control your cloud footprint. You will probably say "I have only 2-3 instances and know exactly what their purpose is", but because it is easy to scale up, you will find yourself reaching to few dozen of instances with a snap of your fingers. From my experience beyond 30 instances things get complicated!
Take Control – To continue the last topic, set rules and procedures to track your resources purchasing
- Maintain an Excel file or a database, and store all the information about your cloud footprint in it from the start
- Document everything including: 'AMI' names, Instance Tag's, Machines' purposes, Application Types, Availability Zones of the Instances, Snap-shots, Elastic IPs and so on
- Make sure that you link the business need to the enabled cloud resource
Select the Right Size –common habit is to take a “one size fits all” approach to your cloud client instances. This might be a fast and convenient solution, but it contributes a lot to overspending. I strongly suggest providing the right size (obviously) for each individual request. Another option is to provide a smaller instance and scale up later if needed. Just to help you understand the cost impact, one hour of an EC2 small Windows instance costs 0.12c while one level above, an EC2 large Windows instance costs 0.48c, four times more!
Use reserved and spot instances - Plan your need for instances and take advantage of AWS capabilities to provision low cost instances (this is how they utilize their unused capacity). I met a company that made a bid for 10 spot instances for 20 cents and they found that they meet the low availability needs of their internal application testing environment. Cool, isn't it? check Amazon AWS calculator
Control your storage costs - Surprise, surprise, you do have storage and not just servers! Any AMI or snap-shot you take, costs money!
- Unused AMI’s can and should be deleted. After creating an instance the AMI is not deleted. Why? the only logic goes that you have a ready image that you can turn into a working one or more instances in a short time
- Take control of those excess snapshots! In AWS you pay for every Snapshot (storage costs)
- Remove duplicate idle and unknown files. The cloud is not your personal laptop, you pay per use and as you go for every byte even if it isn’t in use
Forgive me if I sounded a bit harsh but I find that cost is one of the main reasons why companies go back to their internal IT paradigm which is a shame. The cloud can be great if you know how to utilize it right.
Your Cloud Guy,
I could not agree more. Last week I wrote a similar post on why corporate IT is oblivious to their new role. But, I do think the renaissance of IT is coming. Either that or the same time next week we will be using EC2 dashboard to provision new server ;)
Posted by: Supras On Sale | 11/29/2011 at 09:46 AM